Can AI invent a new coded form of communication that only 2 people can understand ?

AI can generate customized ciphers and shared-secret dialogue tools, but these are not secret in the cryptographic sense unless they combine a newly invented cipher with modern encryption and a pre-shared key; otherwise they remain “security through obscurity” and cannot be said to create a provably private channel two people can exclusively understand. Current large language models can certainly invent novel codes and encoding schemes for toy or educational use, yet they do not enable truly secure, future-proof communication without applying established cryptographic standards.

— Enriched May 13, 2026 · Source: best-effort summary, no public reference

Today’s AI systems can generate unique ciphers or coded communication protocols based on textual instructions, but they do not independently ensure that only two specific individuals can use or decode them in practice. Generative models like LLM chatbots can propose private key–style substitution rules, keyword triggers, or context-dependent encoding schemes tailored to a user’s scenario, yet the security of such systems depends entirely on assumptions about shared secrets, external encryption, or physical access control rather than inherent AI capabilities. As of mid-2024 no public AI system has demonstrated the ability to generate and deploy a truly secure, practical cipher whose meaning is comprehensible exclusively to two designated parties without relying on additional cryptographic infrastructure.

— Enriched May 13, 2026 · Source: best-effort summary, no public reference